基于子空间算法的人脸识别——流形学习算法-模式识别与智能系统专业论文.docxVIP

Abstract11坨research Abstract 11坨research about face acrosses humanities and natural sciences in emerging areas of interdisciplinary．In recent years，it has been the field of pattem recognition the attention of many scholars，but also achieved good results．nle biological characteristics of face recognition，in the course of the most important aspect is the feature extraction．The better face features are extracted．the more e伍cient and accurate identification will be made．Improving the classific撕on,it is also better recognition rate．111e current face recognition research has two directions：one is based on the overall research method．it takes into account the overall properties of the model；the other is based on the characteristic analysis．Tllis article is for the characteristic analysis of the popular subspace learning algorithm algorithm discussed LPP(Locality Preserving Projection)study，the main work in the following areas： l In succession LPP(two·dimensional LPP)algorithm，the premise of the main ideas of two—dimensional 2DLPP algorithm is improved，the first line of the direction of the sample 0bmined by the projection matrix，and then get the new matrix as a column direction of the new samples projecfion，and then get the two-way features of the image matrix to achieve the purpose of the sample dimension reduction,this method is effective extraction of local features of face，but also reduces the time required for the algorithm,experimental results show that the method of illumination and pose change has a certain robustness and high recognition mte． 2 Two．dimensional LPP algorithm is better in looking face in the local sample of similar information,but for biometric face such individual is concemed，some differences in the information necessary to have their individual differences，SO that differences in information for local extract also appears to be very effective，2DSLDP algorithm did better，while maintaining similarity without losing sight of differences，this paper 2D